Declare explicit, receipt-verifiable objective lifecycle relations

[bigboateng]

Summary

• Replace the kernel's overloaded BindObjectiveMutation with five explicit lifecycle relations — preserve, bind-initial, advance, replace, and clear — each requiring its own capability (objective.bind, objective.advance, objective.replace, objective.clear). Initial binds require verified absence and may bind any valid externally versioned revision; advances keep the objective identity and move to exactly the next revision, rejecting gaps, downgrades, duplicates, overflow, and same-revision fingerprint substitution; replacements must be declared and rebind a different identity; clears remove an existing binding; preserve cannot silently change one.
• Record the declared relation in prescriptions (schema 3) and receipts (schema 5); receipt validation proves the relation from the receipt's own prior/requested/result binding lineage without recovering transition semantics from a program body. Invalid lifecycle relations are rejected before effects, and freshness stays bound to exact state, observation, program, authority, and objective binding.
• Replace software delivery's binds_requested_objective policy flag with the declared relation: objective.bind configures the first objective over verified absence, and a new objective.replace core transition declares different-objective reconfiguration, projected through the compiler into the kernel's replace relation. Admissions bump to schema 10 and delivery receipts to schema 14 with exact committed-history readers for the previous shapes and no aliases or shims. Untargeted canonical selection still reconfigures a supplied different objective — now through the declared replacement.
• Add a domain-neutral ObjectiveLifecycleConformance suite over a multi-revision register: valid binds, advances, declared replacements, and clears; rejection of every invalid lineage before effects; per-relation capability isolation; concurrent successors settling exactly once; stale and pre-clear prescriptions; verification and commit failures that never advance accepted state; replay and cross-instance replay rejection; restart reconstruction; and receipt-chain continuity. White-box law tests prove torn commits, blind concurrent commits, cleared-binding resurrection, relabeled relations, and restart state resets fail the shared laws.
• Regenerate the checked catalog artifacts (64 executable transitions), update the repository contract pins, architecture and workflow docs, and add the release note.

Test plan

• go vet ./... && go build ./... (boatstack)
• go test ./... — full suite including the new TestObjectiveLifecycleConformance and dishonest-variant law tests
• go test -race ./... — full race suite
• python3 -m unittest discover -s .github/tests -p 'test_*.py' — repository contract, 69 tests
• npm run test:flow-sdk && npm run docs:check and required-mode frontend conformance (BOATSTACK_REQUIRE_FLOW_FRONTEND=1)
• Release-note policy (release_notes.py check-policy) over the exact range
• Boatstack self-review converged with zero blocking findings; sealed attestation committed